381
Atherosclerosis, 32 (1979) 381-396 0 Elsevier/North-Holland Scientific Publishers, Ltd.
NATURE
OF THE FAMILIAL
INFLUENCE
ON PLASMA LIPID LEVELS
D. HEWITT, G.J.L. JONES, G.J. GODIN and DOREEN WRAIGHT (Epidemiology and Statistics), W.C. BRECKENRIDGE (Loboratory), and J.A. LITTLE, G. STEINER and M.A. MISHKEL (Clinical) University of Toronto, Toronto, Ont. and McMaster Lipid Research Clinic, McMaster University, Hamilton, Ont. (Canada) (Received 10 July, 1978) (Revised, received 21 November, 1978) (Accepted 29 November, 1978) ---
Summary In order to test for the presence of ‘major’ genetic factors influencing plasma concentration of CH and TG a sample of nearly 500 nuclear families, each including father, mother and two or more natural children, was drawn from the general, well population of the Toronto-Hamilton region. Patterns suggestive of segregation were obtained from statistical analysis of plasma concentrations expressed on the scale of mg/lOO ml and weaker, but possibly more valid authentic patterns, from measurements expressed on a logarithmic scale. In particular, CH variance was significantly increased in the progeny of subjects with hyperalphalipoproteinemia and of those with Type II HLP, while TG variance was increased in the progeny of subjects with Type IV HLP. The resulting contribution to total population variance at childhood ages, however, appears to be rather small. An analysis of the same data in terms of correlation coefficients showed that familial factors influencing lipid levels have a fairly high degree of specificity as between CH and TG, and apply in the lower as well as the upper ‘range of plasma concentrations. Familial correlations tend to be stronger for blood relatives than for unrelated members of the same household, stronger for CH than for TG, stronger for mothers than for fathers, and stronger for younger than for older children. Key words:
Cholesterol - Familial - Inheritance -Lipids
Supported by Contract No. NIH-NHLBI No l-HV 2-2917-L, Dept. of HEW, and by an Ontario Heart Foundation Grant.
-Plasma
lipids - Triglyceride
under the Lipid Research Climic Program.
382
Introduction The Prevalence Study carried out by the Toronto-McMaster Lipid Research Clinic [l] is one of a number of population-based studies in which significant resemblance among family members has been demonstrated, with sib-sib and parent-offspring correlations similar to those reported from the Tecumseh study [ 21. Such findings are generally, though guardedly, interpreted as evidence of a genetic influence on lipid levels. However, the importance of genetic factors has been brought into question by a report that, in the communal setting of the Israeli kibbutzim, no special resemblance could be found among the members of biological families [3]. Twin data on lipid levels meet the traditional criterion for a genetic effect - closer similarity of monozygotic than of dizygotic twins -but, when analysed by the method of Christian et al., these same data have been used to support the view that the variance of esterified - as distinct from free cholesterol (specifically in adult males) has no genetic component [ 4,5]. Clinically-based studies have tended to assume the reality of a genetic influence on lipid levels, rather than to demonstrate it, and have argued the plausibility of this or that particular genetic mechanism on somewhat subjective grounds. An exception is provided by the work of Goldstein et al. [ 61, who made use of an explicit and objective criterion for discriminating between patterns characteristic of multifactorial inheritance and patterns characteristic of major, dominant gene effect. If resemblance among relatives is determined by multifactorial inheritance then the within-family variance (measured on some appropriate scale, see Discussion) will be the same for families in which the average level is high as for those in which it is lower. However, if there are single genes that have a major effect then families in which segregation of these genes is occurring will, on average, exhibit a larger within-family variance than families whose members are all “normal”. Subject to some technical reservations discussed below this means that the variance ratio can be used as a test statistic for detecting the presence of major gene effects. One may hypothesize that a trait whose variation is governed solely by environmental influences will, in respect of within-family variance, behave like one governed by multifactorial inheritance. Hence a significantly large variance ratio will serve to discredit two null hypotheses simultaneously: (i) that there are no genetic effects, and (ii) that any genetic effects present are of the multifactorial or “microphenic” [ 71 type only. If the criterion of within-family variance is to be applied in an impartial manner it will be necessary to estimate the within-family variance from a on the basis of proband characteristics alone, the series of families constituted proband being excluded from the variance calculation, and to test whether this exceeds the corresponding estimate for a suitable control series of families. A component of the Toronto-McMaster study was designed specifically for the purpose of carrying out such tests and was completed during 1975-76 under the sponsorship of the Lipid Research Clinic Program [ 81. Adult subjects who came to an initial screen were asked whether they had a spouse and two or more children living in the area who might also be willing to provide blood samples for determination of fasting cholesterol and triglyceride. In this way
383
we were able to define groups of families selected by index subjects who were “normal” or “abnormal” either in terms of lipid level or in terms of lipoprotein constitution, and who belonged to either the parental or the filial generation. Such groups could then be compared on the basis of within-sibship variance (if the index subjects used were parents), or on the basis of variance within spouse pairs (if the index subjects were children). The genetical interpretation of such variance comparisons is deferred to the Discussion Section, but it is necessary at this point to raise a matter concerning the statistical validity of the comparisons and the reason for carrying them out in duplicate. The key assumption underlying application of the variance ratio, or F-statistic is that, if the null hypothesis is correct, lipid values for each member of a family may be regarded as independently drawn from a normallydistributed population of values having the same variance as the population from which any other family’s values are drawn. Some degree of departure from normality will not be critical, but the test would become biassed if population variance is dependent on mean value, because a sample having a higher mean value (relatives of hyperlipidemic subjects) is to be compared with a sample having a lower mean value (relatives of normals). Since it is commonly held that lipid level is more variable in subjects whose average values over a period of time are high it might seem prudent to compensate for this by a transformation of the data, such as taking the logarithms of the raw or age-adjusted values, and this has been done in the present study. On the other hand, if the null hypothesis is incorrect and major gene effects do operate, then the use of a log transformation will tend to suppress the very phenomenon we want to detect, by forcing the values for members of an abnormal, high-lipid sub-population towards or into the body of the frequency distribution for normals. The same consideration applies to any other normalizing transformation, such as the power transform used by Morton et al. [ 71. Under these circumstances the use of transformed data alone would be unduly conservative, so tests based on untransformed data will also be reported below. In the case of within-spouse-pair variance the key assumption for application of the F-test is not met, since adult male populations tend to have somewhat larger variance for lipid measurements than do female populations. It might be possible to make an appropriate scale adjustment, over and above the adjustment to a common mean, but this has not been done with the TorontoMcMaster data. Certain of the test results below, specifically those related to Table 2, are accordingly subject to some bias favoring rejection of the null hypothesis. This may be of some importance in relation to the untransformed triglyceride values. Besides reporting on the outcome of the tests for which the family component of the Toronto-McMaster Prevalence Study was designed, the present paper also describes some more exploratory and descriptive analyses that were prompted by the results obtained. Methods Description of the target population of the Toronto-McMaster the sampling procedure, the method of drawing blood, preparing
lipid survey, and analyzing
384
the blood samples and the quality control system can be found elsewhere [g-12]. Briefly, a total of 6409 working adults in the Toronto-Hamilton area .were tested for fasting plasma cholesterol and triglyceride levels between 1973 and 1975. Subsequently 1248 children and 520 spouses of those married workers who had two or more children were also screened. These families form the study series for the present paper. Though not a random sample of the main series they were selected without reference to the participating parents’ lipid level. About 25% of the participants of the initial screen (Visit l), including all of those above the NHLBI 90th percentile for age-specific cholesterol or the 95th percentile for trigiytieride, were asked to attend a second, more detailed visit in a hospital clinic setting (Visit 2). At Visit 2 a larger sample of blood was drawn to permit measurement of the cholesterol concentrations of the HDL, ,LDL and VLDL lipoproteins, since it was expected that most cases of diagnosable lipoprotein abnormality would lie in the upper 5% or JO% of the total CH or TG distribution. Frozen plasma and serum from each Visit 2 participant were sent to the Central Clinical Chemistry Laboratory (Bio-Science Labs in Van Nuys, California) for measurement of plasma glucose, serum alkaline phosphatase, total bilirubin, total protein, creatinine, uric acid, SGOT, T4 thyroxine [13]. These tests were used to identify possible causes of secondary hyperlipoproteinemia. Visit 1 and Visit 2 blood specimens were processed at the lipid laboratory in .Toronto and lipids analysed according to procedures specified in the LRC Laboratory Methods Manual [12]. Total plasma cholesterol and triglyceride determinations were made using Technicon Auto-Analyzer II methodologies, adapted by the LRC ‘program. Lipoproteins were separated by ultracentrifugal flotation (40,000 RPM, 18 h on a Beckman Rotor 40.3) at saline density (D k.l.006 g/ml) to yield a supernatant fraction containing VLDL and an infranatant fraction containing LDL and HDL. HDL was estimated in total plasma following the precipitation of heparin and manganese chloride. Following direct estimation of VLDL and HDL, LDL was determined by the formula: LDL cholesterol = cholesterol in 1.006 infranatant - HDL cholesterol. VLDL was estimated from the formula: VLDL = plasma cholesterol - 1.006 infranatant. In cases of incomplete precipitation of VLDL and LDL the procedure was repeated on the infranatant fraction following ultracentrifugation. The presence of chylomicrons was assessed by observing plasma after standing overnight at 4°C. The presence of floating P-lipoprotein was observed when agarose gel electrophoresis of the VLDL fraction revealed lipoproteins with electrophoretic’*mobility equal to &lipoproteins. The presence of sinking pre-P-lipoprotein was observed when the electrophoretic pattern of the infranatant frac,tion contained lipoproteins of pre-fl mobility. The average ages of the parents and of the children were 40 yrs and 11 yrs respectively, with no perceptible differences between the family groups differentiated on the basis of lipid status. Among the children 52% were boys, again with no perceptible differences from group to group. Hence no adjustment for age or sex would have been needed in order to ensure comparability of these ‘groups; but adjustments were made nevertheless in order to remove some of the
385
impermanent interindividual variance. In the case of CH levels this was achieved by fitting quadratic age-trends for each sex separately and then adding to or subtracting from each raw value the predicted difference between average CH at the subject’s age and at age 22. Since the fitted age trends for males and females intersected at age 22 this was effectively an adjustment for sex also. For the analyses presented in Table 1 below, the In CH values were logarithms of the age-adjusted CH values. TG values were adjusted in a similar manner except that a linear rather than a quadratic trend was fitted to the data for females, and that In TG values were obtained by taking logarithms before rather than after fitting the age trend. It should be understood that while these procedures bring the data for both sexes and all ages to a common mean they do not in any way standardize the variance. Thirteen families were regularly excluded because of laboratory evidence that the HLP affecting a family member might be secondary to some other disease process. Among factors other than age, sex and illness that are known or believed to affect lipid status (e.g. race, socio-economic status, skinfold, exogenous hormones) none were considered important enough in the study population to warrant special adjustment or control. In some families measurements were obtained on only one parent or only one eligible child. These families were sometimes included and sometimes excluded from the study, depending on the requirements of the particular statistical analysis. No use has been made in the present study of stepchildren, half-sibs, or twin-born offspring. Results Table 1 reports a series of tests of the major gene hypothesis (versus the null hypotheses of no genetical influence and of multifactorial influence alone) in which groups of “normal” or potentially “abnormal” sibships were defined on the basis of the lipid status of the parents. For tests reported in the upper half of the Table parents’ lipid status was classified in terms of total plasma cholesterol (CH) and triglyceride (TG) as measured at Visit 1. It will be seen that elevation of one or both of these levels in one or both parents was found in 71 families containing 182 (=71 + 111) children, and that within these families the variance of CH was 15% above the control level of 428.0 (not a significant margin), while the variance of TG was 40% above the control level of 641.5 (P < 0.01). Restricting attention to families in which one or both parents showed elevation of CH a significant (P < 0.05) increase of within sibship variance was found for both CH and TG. In the 29 families (with 82 children) where one or both parents had elevated TG, only the variance of TG was increased. When these tests were repeated with CH and TG transformed to a log scale (see right hand side of Table 1) there was little or no evidence of increased variance in the sibships from “abnormal” families and none of the results were statistically significant. For tests reported in the lower half of Table 1 parents’ lipid status was classified according to the Fredrickson-Levy-Lees typology [11,14]. The 14 families in which a parent was classified as having a Type II HLP constitute just over 3% of the study series (approximately 14% of the individual parents being
78.3 75.5
168.1
171.4 161.1
71
49 29
a P < 0.05. b P < 0.01.
Both free of lipoprotein abnormaUty Any lipoprotein abnormality Type 11s or IIb Type III Type IV Hyperalpha Other or multiple abnormality
76.5
154.8
351
Both within normal range of CH and TG Elevation of CH or TG in one or both Elevation of CH Elevation of TG
154.8
166.2
176.6 180.0 160.6 192.7 161.9
350
76
14 2 36 4 20
CH
76.2 70.5 79.0 62.7 79.7
77.6
70.8
70.8
TG
25 2 61 5 30
121
590
79 53
111
586
WHOSE
PARENTS
742.1 20.7 316.1 1533.6 432.4
483.3
428.6
566.6 413.4
491.5
1.73 a
Atherosclerosis, 32 (1979) 381-396 0 Elsevier/North-Holland Scientific Publishers, Ltd.
NATURE
OF THE FAMILIAL
INFLUENCE
ON PLASMA LIPID LEVELS
D. HEWITT, G.J.L. JONES, G.J. GODIN and DOREEN WRAIGHT (Epidemiology and Statistics), W.C. BRECKENRIDGE (Loboratory), and J.A. LITTLE, G. STEINER and M.A. MISHKEL (Clinical) University of Toronto, Toronto, Ont. and McMaster Lipid Research Clinic, McMaster University, Hamilton, Ont. (Canada) (Received 10 July, 1978) (Revised, received 21 November, 1978) (Accepted 29 November, 1978) ---
Summary In order to test for the presence of ‘major’ genetic factors influencing plasma concentration of CH and TG a sample of nearly 500 nuclear families, each including father, mother and two or more natural children, was drawn from the general, well population of the Toronto-Hamilton region. Patterns suggestive of segregation were obtained from statistical analysis of plasma concentrations expressed on the scale of mg/lOO ml and weaker, but possibly more valid authentic patterns, from measurements expressed on a logarithmic scale. In particular, CH variance was significantly increased in the progeny of subjects with hyperalphalipoproteinemia and of those with Type II HLP, while TG variance was increased in the progeny of subjects with Type IV HLP. The resulting contribution to total population variance at childhood ages, however, appears to be rather small. An analysis of the same data in terms of correlation coefficients showed that familial factors influencing lipid levels have a fairly high degree of specificity as between CH and TG, and apply in the lower as well as the upper ‘range of plasma concentrations. Familial correlations tend to be stronger for blood relatives than for unrelated members of the same household, stronger for CH than for TG, stronger for mothers than for fathers, and stronger for younger than for older children. Key words:
Cholesterol - Familial - Inheritance -Lipids
Supported by Contract No. NIH-NHLBI No l-HV 2-2917-L, Dept. of HEW, and by an Ontario Heart Foundation Grant.
-Plasma
lipids - Triglyceride
under the Lipid Research Climic Program.
382
Introduction The Prevalence Study carried out by the Toronto-McMaster Lipid Research Clinic [l] is one of a number of population-based studies in which significant resemblance among family members has been demonstrated, with sib-sib and parent-offspring correlations similar to those reported from the Tecumseh study [ 21. Such findings are generally, though guardedly, interpreted as evidence of a genetic influence on lipid levels. However, the importance of genetic factors has been brought into question by a report that, in the communal setting of the Israeli kibbutzim, no special resemblance could be found among the members of biological families [3]. Twin data on lipid levels meet the traditional criterion for a genetic effect - closer similarity of monozygotic than of dizygotic twins -but, when analysed by the method of Christian et al., these same data have been used to support the view that the variance of esterified - as distinct from free cholesterol (specifically in adult males) has no genetic component [ 4,5]. Clinically-based studies have tended to assume the reality of a genetic influence on lipid levels, rather than to demonstrate it, and have argued the plausibility of this or that particular genetic mechanism on somewhat subjective grounds. An exception is provided by the work of Goldstein et al. [ 61, who made use of an explicit and objective criterion for discriminating between patterns characteristic of multifactorial inheritance and patterns characteristic of major, dominant gene effect. If resemblance among relatives is determined by multifactorial inheritance then the within-family variance (measured on some appropriate scale, see Discussion) will be the same for families in which the average level is high as for those in which it is lower. However, if there are single genes that have a major effect then families in which segregation of these genes is occurring will, on average, exhibit a larger within-family variance than families whose members are all “normal”. Subject to some technical reservations discussed below this means that the variance ratio can be used as a test statistic for detecting the presence of major gene effects. One may hypothesize that a trait whose variation is governed solely by environmental influences will, in respect of within-family variance, behave like one governed by multifactorial inheritance. Hence a significantly large variance ratio will serve to discredit two null hypotheses simultaneously: (i) that there are no genetic effects, and (ii) that any genetic effects present are of the multifactorial or “microphenic” [ 71 type only. If the criterion of within-family variance is to be applied in an impartial manner it will be necessary to estimate the within-family variance from a on the basis of proband characteristics alone, the series of families constituted proband being excluded from the variance calculation, and to test whether this exceeds the corresponding estimate for a suitable control series of families. A component of the Toronto-McMaster study was designed specifically for the purpose of carrying out such tests and was completed during 1975-76 under the sponsorship of the Lipid Research Clinic Program [ 81. Adult subjects who came to an initial screen were asked whether they had a spouse and two or more children living in the area who might also be willing to provide blood samples for determination of fasting cholesterol and triglyceride. In this way
383
we were able to define groups of families selected by index subjects who were “normal” or “abnormal” either in terms of lipid level or in terms of lipoprotein constitution, and who belonged to either the parental or the filial generation. Such groups could then be compared on the basis of within-sibship variance (if the index subjects used were parents), or on the basis of variance within spouse pairs (if the index subjects were children). The genetical interpretation of such variance comparisons is deferred to the Discussion Section, but it is necessary at this point to raise a matter concerning the statistical validity of the comparisons and the reason for carrying them out in duplicate. The key assumption underlying application of the variance ratio, or F-statistic is that, if the null hypothesis is correct, lipid values for each member of a family may be regarded as independently drawn from a normallydistributed population of values having the same variance as the population from which any other family’s values are drawn. Some degree of departure from normality will not be critical, but the test would become biassed if population variance is dependent on mean value, because a sample having a higher mean value (relatives of hyperlipidemic subjects) is to be compared with a sample having a lower mean value (relatives of normals). Since it is commonly held that lipid level is more variable in subjects whose average values over a period of time are high it might seem prudent to compensate for this by a transformation of the data, such as taking the logarithms of the raw or age-adjusted values, and this has been done in the present study. On the other hand, if the null hypothesis is incorrect and major gene effects do operate, then the use of a log transformation will tend to suppress the very phenomenon we want to detect, by forcing the values for members of an abnormal, high-lipid sub-population towards or into the body of the frequency distribution for normals. The same consideration applies to any other normalizing transformation, such as the power transform used by Morton et al. [ 71. Under these circumstances the use of transformed data alone would be unduly conservative, so tests based on untransformed data will also be reported below. In the case of within-spouse-pair variance the key assumption for application of the F-test is not met, since adult male populations tend to have somewhat larger variance for lipid measurements than do female populations. It might be possible to make an appropriate scale adjustment, over and above the adjustment to a common mean, but this has not been done with the TorontoMcMaster data. Certain of the test results below, specifically those related to Table 2, are accordingly subject to some bias favoring rejection of the null hypothesis. This may be of some importance in relation to the untransformed triglyceride values. Besides reporting on the outcome of the tests for which the family component of the Toronto-McMaster Prevalence Study was designed, the present paper also describes some more exploratory and descriptive analyses that were prompted by the results obtained. Methods Description of the target population of the Toronto-McMaster the sampling procedure, the method of drawing blood, preparing
lipid survey, and analyzing
384
the blood samples and the quality control system can be found elsewhere [g-12]. Briefly, a total of 6409 working adults in the Toronto-Hamilton area .were tested for fasting plasma cholesterol and triglyceride levels between 1973 and 1975. Subsequently 1248 children and 520 spouses of those married workers who had two or more children were also screened. These families form the study series for the present paper. Though not a random sample of the main series they were selected without reference to the participating parents’ lipid level. About 25% of the participants of the initial screen (Visit l), including all of those above the NHLBI 90th percentile for age-specific cholesterol or the 95th percentile for trigiytieride, were asked to attend a second, more detailed visit in a hospital clinic setting (Visit 2). At Visit 2 a larger sample of blood was drawn to permit measurement of the cholesterol concentrations of the HDL, ,LDL and VLDL lipoproteins, since it was expected that most cases of diagnosable lipoprotein abnormality would lie in the upper 5% or JO% of the total CH or TG distribution. Frozen plasma and serum from each Visit 2 participant were sent to the Central Clinical Chemistry Laboratory (Bio-Science Labs in Van Nuys, California) for measurement of plasma glucose, serum alkaline phosphatase, total bilirubin, total protein, creatinine, uric acid, SGOT, T4 thyroxine [13]. These tests were used to identify possible causes of secondary hyperlipoproteinemia. Visit 1 and Visit 2 blood specimens were processed at the lipid laboratory in .Toronto and lipids analysed according to procedures specified in the LRC Laboratory Methods Manual [12]. Total plasma cholesterol and triglyceride determinations were made using Technicon Auto-Analyzer II methodologies, adapted by the LRC ‘program. Lipoproteins were separated by ultracentrifugal flotation (40,000 RPM, 18 h on a Beckman Rotor 40.3) at saline density (D k.l.006 g/ml) to yield a supernatant fraction containing VLDL and an infranatant fraction containing LDL and HDL. HDL was estimated in total plasma following the precipitation of heparin and manganese chloride. Following direct estimation of VLDL and HDL, LDL was determined by the formula: LDL cholesterol = cholesterol in 1.006 infranatant - HDL cholesterol. VLDL was estimated from the formula: VLDL = plasma cholesterol - 1.006 infranatant. In cases of incomplete precipitation of VLDL and LDL the procedure was repeated on the infranatant fraction following ultracentrifugation. The presence of chylomicrons was assessed by observing plasma after standing overnight at 4°C. The presence of floating P-lipoprotein was observed when agarose gel electrophoresis of the VLDL fraction revealed lipoproteins with electrophoretic’*mobility equal to &lipoproteins. The presence of sinking pre-P-lipoprotein was observed when the electrophoretic pattern of the infranatant frac,tion contained lipoproteins of pre-fl mobility. The average ages of the parents and of the children were 40 yrs and 11 yrs respectively, with no perceptible differences between the family groups differentiated on the basis of lipid status. Among the children 52% were boys, again with no perceptible differences from group to group. Hence no adjustment for age or sex would have been needed in order to ensure comparability of these ‘groups; but adjustments were made nevertheless in order to remove some of the
385
impermanent interindividual variance. In the case of CH levels this was achieved by fitting quadratic age-trends for each sex separately and then adding to or subtracting from each raw value the predicted difference between average CH at the subject’s age and at age 22. Since the fitted age trends for males and females intersected at age 22 this was effectively an adjustment for sex also. For the analyses presented in Table 1 below, the In CH values were logarithms of the age-adjusted CH values. TG values were adjusted in a similar manner except that a linear rather than a quadratic trend was fitted to the data for females, and that In TG values were obtained by taking logarithms before rather than after fitting the age trend. It should be understood that while these procedures bring the data for both sexes and all ages to a common mean they do not in any way standardize the variance. Thirteen families were regularly excluded because of laboratory evidence that the HLP affecting a family member might be secondary to some other disease process. Among factors other than age, sex and illness that are known or believed to affect lipid status (e.g. race, socio-economic status, skinfold, exogenous hormones) none were considered important enough in the study population to warrant special adjustment or control. In some families measurements were obtained on only one parent or only one eligible child. These families were sometimes included and sometimes excluded from the study, depending on the requirements of the particular statistical analysis. No use has been made in the present study of stepchildren, half-sibs, or twin-born offspring. Results Table 1 reports a series of tests of the major gene hypothesis (versus the null hypotheses of no genetical influence and of multifactorial influence alone) in which groups of “normal” or potentially “abnormal” sibships were defined on the basis of the lipid status of the parents. For tests reported in the upper half of the Table parents’ lipid status was classified in terms of total plasma cholesterol (CH) and triglyceride (TG) as measured at Visit 1. It will be seen that elevation of one or both of these levels in one or both parents was found in 71 families containing 182 (=71 + 111) children, and that within these families the variance of CH was 15% above the control level of 428.0 (not a significant margin), while the variance of TG was 40% above the control level of 641.5 (P < 0.01). Restricting attention to families in which one or both parents showed elevation of CH a significant (P < 0.05) increase of within sibship variance was found for both CH and TG. In the 29 families (with 82 children) where one or both parents had elevated TG, only the variance of TG was increased. When these tests were repeated with CH and TG transformed to a log scale (see right hand side of Table 1) there was little or no evidence of increased variance in the sibships from “abnormal” families and none of the results were statistically significant. For tests reported in the lower half of Table 1 parents’ lipid status was classified according to the Fredrickson-Levy-Lees typology [11,14]. The 14 families in which a parent was classified as having a Type II HLP constitute just over 3% of the study series (approximately 14% of the individual parents being
78.3 75.5
168.1
171.4 161.1
71
49 29
a P < 0.05. b P < 0.01.
Both free of lipoprotein abnormaUty Any lipoprotein abnormality Type 11s or IIb Type III Type IV Hyperalpha Other or multiple abnormality
76.5
154.8
351
Both within normal range of CH and TG Elevation of CH or TG in one or both Elevation of CH Elevation of TG
154.8
166.2
176.6 180.0 160.6 192.7 161.9
350
76
14 2 36 4 20
CH
76.2 70.5 79.0 62.7 79.7
77.6
70.8
70.8
TG
25 2 61 5 30
121
590
79 53
111
586
WHOSE
PARENTS
742.1 20.7 316.1 1533.6 432.4
483.3
428.6
566.6 413.4
491.5
1.73 a
